* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download pdf
Internet protocol suite wikipedia , lookup
Deep packet inspection wikipedia , lookup
Zero-configuration networking wikipedia , lookup
Distributed firewall wikipedia , lookup
Wake-on-LAN wikipedia , lookup
Computer network wikipedia , lookup
Backpressure routing wikipedia , lookup
Distributed operating system wikipedia , lookup
Cracking of wireless networks wikipedia , lookup
List of wireless community networks by region wikipedia , lookup
Multiprotocol Label Switching wikipedia , lookup
Airborne Networking wikipedia , lookup
Recursive InterNetwork Architecture (RINA) wikipedia , lookup
Spanning Tree Protocol wikipedia , lookup
Dijkstra's algorithm wikipedia , lookup
CS 5114 Network Programming Languages Control Plane http://www.flickr.com/photos/rofi/2097239111/ Nate Foster Cornell University Spring 2013 Based on lecture notes by Jennifer Rexford and Michael Freedman Announcements Breakfast pickup and presentations posted to website Reviews: start today! • • Only review one paper but please read them all Please submit by CMS in the future Homework #1 • • • Will go out next Tuesday Due 2 weeks later Topic: OpenFlow programming Overview Host (last time) • • Network discovery and bootstrapping Resource allocation and interface to applications Data plane (next time) • • Streaming algorithms and switch fabric Forward, filter, buffer, schedule, mark, monitor, … Control plane (today) • • Distributed algorithms for computing paths Disseminating the addresses of end hosts Data, Control, and Management Planes Data Control Management Timescale Packet (ns) Event (10 ms to sec) Human (min to hours) Tasks Forwarding, buffering, filtering, scheduling Routing, signaling Analysis, configuration Location Line-card hardware Router software Humans or scripts Routing vs. Forwarding Routing: control plane § Computing paths the packets will follow § Routers talking amongst themselves § Individual router creating a forwarding table Forwarding: data plane § Directing a data packet to an outgoing link § Individual router using a forwarding table Data and Control Planes control plane data plane Processor Line card Line card Line card Line card Switching Fabric Line card Line card 6 Routing Protocols What does the protocol compute? § Spanning tree, shortest path, local policy, arbitrary endto-end paths? What algorithm does the protocol run? § Spanning-tree construction, distance vector, link-state routing, path-vector routing, source routing, end-toend signaling How do routers learn end-host locations? § Learning/flooding, injecting into the routing protocol, dissemination using a different protocol, and directory server What Does the Protocol Compute? Different Ways to Represent Paths Static Model The outcome of routing computations § Not how the (distributed) computations are performed § Trade-offs State required to represent the paths § Efficiency of the resulting paths § Ability to support multiple paths § Complexity of computing paths § Which nodes control the computation § Different Settings § LAN, intradomain, interdomain Spanning Tree A tree that connects every node Single path between each pair of nodes § No loops, so supports broadcast easily § Disadvantages Paths can sometimes be long § Some links unused! § Shortest Paths Shortest path(s) between each pair of nodes Separate shortest-path tree rooted at each node § Minimum hop count (or minimum sum of weights) § Disadvantages All nodes must agree on the link metrics § Multipath routing is limited (e.g., Equal Cost Multipath) § Local Policy at Each Hop Locally best path Each node picks the path it likes best § … from among the paths selected by its neighbors § Disadvantages § More complicated to configure and model 21d 2 2d 32d 34d 3 54d 5 4d 1d 1 12d 4 64d d 6 654d 32d 34d 3 54d 5 2 4d 21d 2d 1d 1 12d 4 64d d 6 654d End-to-End Path Selection End-to-end path selection Each node picks its own end to end paths § … independent of what other paths other nodes use § Disadvantages More state and complexity in the nodes § Hop-by-hop destination-based forwarding is not enough § How to Compute Paths? Spanning Tree Algorithm Elect a root § Select switch with the smallest identifier and form a tree Algorithm § Repeatedly talk to neighbors root – “I think node Y is the root” – “My distance from Y is d” § Update state based on neighbors – Smaller id as the root – Smaller distance d+1 § One hop Disable interfaces not on path Primarily used in Ethernet-based LANs Three hops Spanning Tree Example: Switch #4 Switch #4 thinks it is the root § Sends (4, 0, 4) message to 2 and 7 root Switch #4 hears from #2 1 Receives (2, 0, 2) message from 2 § … and thinks that #2 is the root § And realizes it is just one hop away § Switch #4 hears from #7 Receives (2, 1, 7) from 7 § And realizes this is a longer path § So, prefers its own one-hop path § And removes 4-7 link from the tree § 3 5 2 4 7 6 Shortest-Path Problem Compute: path costs to all nodes From a given source u to all other nodes § Cost of the path through each outgoing link § Next hop along the least-cost path to s § 2 3 u 2 6 1 1 4 1 5 4 3 s Link State: Dijkstra’s Algorithm • Flood the topology information to all nodes • Each node computes shortest paths to other nodes Initialization S = {u} for all nodes v if (v is adjacent to u) D(v) = c(u,v) else D(v) = ∞ Loop add w with smallest D(w) to S update D(v) for all adjacent v: D(v) = min{D(v), D(w) + c(w,v)} until all nodes are in S Used in OSPF and IS-IS Link-State Routing Example 2 3 2 1 1 2 5 4 1 1 1 4 1 5 3 2 3 2 3 4 1 1 1 4 5 1 4 3 2 2 3 4 1 3 2 1 4 4 5 3 Link-State Routing Example (continued) 2 3 2 1 1 2 5 4 1 1 1 4 1 5 3 2 3 2 3 4 1 1 1 4 5 1 4 3 2 2 3 4 1 3 2 1 4 4 5 3 Link State: Shortest-Path Tree Shortest-path tree from u 2 v 3 u 1 2 1 w Forwarding table at u y link 1 4 x 5 4 s z t 3 v w x y z s t (u,v) (u,w) (u,w) (u,v) (u,v) (u,w) (u,w) Distance Vector: Bellman-Ford Algorithm Define distances at each node x § dx(y) = cost of least-cost path from x to y Update distances based on neighbors § dx(y) = min {c(x,v) + dv(y)} over all neighbors v 2 v 3 u 1 2 1 w y 1 4 x 5 4 s z t 3 du(z) = min{c(u,v) + dv(z), c(u,w) + dw(z)} Used in RIP and EIGRP Distance Vector: Count to Infinity Link cost changes: • Good news travels fast • Bad news travels slow: “count to infinity” problem! 60 4 X Y 50 1 Z algorithm continues on! Path-Vector Routing Extension of distance-vector routing Support flexible routing policies § Avoid count-to-infinity problem § Key idea: advertise the entire path Distance vector: send distance metric per dest d § Path vector: send the entire path for each dest d § “d: path (2,1)” 3 “d: path (1)” 1 2 data traffic Used in BGP data traffic d Path-Vector: Faster Loop Detection Node can easily detect a loop Look for its own node identifier in the path § E.g., node 1 sees itself in the path “3, 2, 1” § Node can simply discard paths with loops § E.g., node 1 simply discards the advertisement “d: path (2,1)” 3 “d: path (1)” 2 “d: path (3,2,1)” 1 Path-Vector: Flexible Policies Each node can apply local policies Path selection: Which path to use? § Path export: Which paths to advertise? § Examples Node 2 may prefer the path “2, 3, 1” over “2, 1” § Node 1 may not let node 3 hear the path “1, 2” § 2 3 1 2 3 1 End-to-End Signaling Establish end-to-end path in advance Learn the topology (as in link-state routing) § End host or router computes and signals a path § Routers supports virtual circuits Signaling: install entry for each circuit at each hop § Forwarding: look up the circuit id in the table § 1 1: 7 2: 7 link 7 1: 14 2: 8 link 14 2 link 8 Used in MPLS with RSVP Source Routing Similar to end-to-end signaling But the data packet carries the hops in the path § … rather than the routers storing big tables § End-host control Tell the end host the topology § Let the end host select the end-to-end path § Variations of source routing Strict: specify every hop § Loose: specify intermediate points § Used in IP source routing (but almost always disabled) Learning Where the Hosts Are Finding the Hosts Building a forwarding table § Computing paths between network elements § … and figuring out where the end-hosts are § … to map a destination address to an outgoing link How to find the hosts? § Learning/flooding § Injecting into routing protocol § Dissemination via different protocol § Directory service Learning and Flooding When a frame arrives § Inspect the source address § Associate address with the incoming interface When the frame has an unfamiliar destination § Forward out all interfaces § … except for the one where the frame arrived B B C A Switch learns how to reach A. D A C When in doubt, shout! Used in Ethernet LANs D Inject into Routing Protocol Treat the end host (or subnet) as a node And disseminate in the routing protocol § E.g., flood information about where addresses attach § 2 3 u 2 6 1 1 1 4 4 5 3 s Used in OSPF and IS-IS, especially in enterprise networks ... Disseminate With Another Protocol Distribute using another protocol § One router learns the route § … and shares the information with other routers disseminate route to other routers learn a route to d (e.g., via BGP) Internal BGP (iBGP) used in backbone networks Directory Service Contact a service to learn the location Lookup the end-host or subnet address § … and learn the label to put on the packet § … to get the traffic to the right egress point § directory e “Host d is at egress e” Used in some data centers d s i Encapsulate packet to send to egress e. Conclusion Routing is challenging Distributed computation § Challenges with scalability and dynamics § Many different solutions for different environments § Ethernet LAN: spanning tree, MAC learning, flooding § Enterprise: link-state, inject subnet addresses § Backbone: link-state inside, path-vector routing with neighboring domains, and iBGP dissemination § Data centers: many different solutions, still in flux – E.g., link-state routing or multiple spanning trees – E.g., directory service, inject subnet “Design Philosophy of the DARPA Internet Protocols” (ACM SIGCOMM, 1988) David Clark Design Goals Primary goal § Effective technique for multiplexed utilization of existing interconnected networks (e.g., ARPAnet, packet radio) Important goals Survivability in the face of failure § Multiple types of communication service § Wide variety of network technologies § Less important goals Distributed management of resources § Cost effectiveness § Host attachment with low level of effort § Accountability of resources § Consequences of the Goals Effective multiplexed utilization of existing networks § Packet switching, not circuit switching Continued communication despite network failures § § Routers don’t store state about ongoing transfers End hosts provide key communication services Support for multiple types of communication service § Multiple transport protocols (e.g., TCP and UDP) Accommodation of a variety of different networks § § Simple, best-effort packet delivery service Packets may be lost, corrupted, or delivered out of order Distributed management of network resources § § Multiple institutions managing the network Intradomain and interdomain routing protocols Questions What if we started with different goals? Network management § Less concern about backwards compatibility § More concern about security § Can we address new challenges Management, security, privacy, sensor nets, … § Without sacrificing the other goals? § Without a major change to the architecture? § “End-to-End Routing Behavior in the Internet” (ACM SIGCOMM, 1996; ToN, 1997) Vern Paxson Measurement With Traceroute Traceroute tool to measure the forwarding path Send packets with TTL=1, 2, 3… § Record the source of the “time exceeded” message § TTL=1 Time exceeded source destination TTL=2 Useful, but introduces many challenges Path changes § Non-participating nodes § Inaccurate, two-way measurements § Questions Why can’t we measure the Internet more directly? § What can we do about it? Right division of labor between host and network? For path selection § For network monitoring § How do we fix these routing problems? In a decentralized, federated network § How to incentivize better network management § Backup Slides on Paxson Paper Paxson Study: Forwarding Loops Forwarding loop § Packet returns to same router multiple times May cause traceroute to show a loop If loop lasted long enough § So many packets traverse the loopy path § Traceroute may reveal false loops Path change that leads to a longer path § Causing later probe packets to hit same nodes § Heuristic solution § Require traceroute to return same path 3 times Paxson Study: Causes of Loops Transient vs. persistent Transient: routing-protocol convergence § Persistent: likely configuration problem § Challenges Appropriate time boundary between the two? § What about flaky equipment going up and down? § Determining the cause of persistent loops? § Anecdote on recent study of persistent loops Provider has static route for customer prefix § Customer has default route to the provider § Paxson Study: Path Fluttering Rapid changes between paths Multiple paths between a pair of hosts § Load balancing policies inside the network § Packet-based load balancing Round-robin or random § Multiple paths for packets in a single flow § Flow-based load balancing Hash of some fields in the packet header § E.g., IP addresses, port numbers, etc. § To keep packets in a flow on one path § Paxson Study: Routing Stability Route prevalence Likelihood of observing a particular route § Relatively easy to measure with sound sampling § Poisson arrivals see time averages (PASTA) § Most host pairs have a dominant route § Route persistence How long a route endures before a change § Much harder to measure through active probes § Look for cases of multiple observations § Typical host pair has path persistence of a week § Paxson Study: Route Asymmetry Hot-potato routing Other causes § Customer B § Provider B Asymmetric link weights in intradomain routing Cold-potato routing, where AS requests traffic enter at particular place Consequences § multiple peering points § Early-exit routing Provider A Customer A Lots of asymmetry One-way delay is not necessarily half of the round-trip time